cre[ai]tion
See your creative process, not just the output.
cre[ai]tion is an AI image workspace for designers that keeps every image, prompt, branch, and transformation connected.
I designed how creative workflows are captured and reused, how references are decomposed into editable parts, and how the AI agent communicates what it is doing.
AI can generate quickly, the hard part is keeping track of design traces.
Most AI image tools treat each generation as an isolated result. Once designers explore multiple prompts, references, and directions, the process becomes difficult to compare, retrace, or reuse.
make the creative process visible.
Every generated result stays connected to the inputs and decisions that produced it. Designers can branch from any step, combine directions, and compare alternatives without losing the path that led there. The canvas becomes a visual history of the exploration.
from one board to a workflow you can run again.
every image remembers its parent
One sofa becomes a family of chairs. Every branch stays attached to where it started, so you can compare directions side by side and go back to any step.
design once, run every season.
A fashion team rarely needs a single image. It needs a repeatable set of outputs: front view, detail shots, fabric studies, and on-model imagery.
I designed a way to capture the part of a board that worked and reuse it with new inputs, like a different set of clothes.
saving a workflow
Designers select the successful portion of the board, define which images are inputs and which are outputs, and save the sequence as a reusable workflow. Prompts remain attached to the images they generated, so each step stays inspectable and editable instead of disappearing into a black box.
change the input, keep the workflow
When a new product is dropped into the workflow, the saved sequence regenerates the required views using the same creative logic. The designer can reuse the system without rebuilding the process from scratch.
break a reference into parts you can actually work with
Inspiration often arrives as a complete look, while a designer may only want the silhouette, bag, material, or styling cue. Decompose separates a reference into usable components while preserving the connection to the original image, so the source of each idea remains visible.
make autonomous AI behavior legible
When an agent is working in the background, silence creates uncertainty: is it thinking, waiting, or stuck? I turned the brackets in the cre[ai]tion identity into a lightweight status character that communicates progress, signals when user input is needed, and confirms completion. The goal was to make agent state understandable without adding a heavy system UI.
keep ideation and iteration in one continuous space
Designers can collect references, add a sketch, generate a render, and reshape the result directly on the canvas instead of repeatedly rewriting prompts. Each version remains connected to the previous step, so the interface supports visual iteration rather than forcing the user back into a text-only workflow.
taste report using claude skills
I also built a Claude skill that turns a list of runway shows into a sourced, bilingual trend report using our team’s template. The workflow separates sourced facts from our own interpretation and links every factual claim back to its source. This let me use AI for speed without giving up traceability or editorial judgment.
faster to finish, faster to run again.
higher task completion after the core generation experience was redesigned as one multimodal canvas.
less repeat setup time, from saving a working setup as a workflow and rerunning it with new inputs.
beta testers the new canvas was measured with.
from an image to more than an image.
Take a generated image into 3D, so an idea can be turned around and checked as an object.
Turn a still into motion for campaigns and ads.
Copyright © 2026 Anqi Wang